The scalar_group helper currently takes first/last
scalar objects, and assumes that everything in
between is valid. This is similar to the default
behaviour of the table helper w.r.t column objects.
However, the table helper can also use a linked list
of range structures, to support "missing columns" -
i.e. non-consecutive ranges of valid column objects.
This is not possible with scalar groups.
It would be useful to extend the scalar group helper
to offer similar functionality (probably using the
same range structures). It might even be possible
to share some of the same code for this processing.